Names binding scope type checking Plan of this

Names, binding, scope, type checking Plan of this part of the course • Names, variables • Binding • Scope • Constants. Variable initialization • Type checking 1

Names and variables Points to discuss • What needs a name? • A variable is a six-tuple 2

Names • A name is a handle on an entity in a program. We refer to something by a name if we want to create, use, change, destroy it. • Declaration, use, lifetime, scope of names: these properties are a major consideration in programming languages. 3

What needs a name? • constants, variables, • operators, • statement labels, • types, • procedures, functions, methods, • modules, programs, • files, disks, • commands, menu items, • computers, networks, user (login name). 4

Names and identifiers • A name is denoted by an identifier. • Typical identifiers are built of letters, digits and underscores, but some languages allow "funny" identifiers as well. Prolog, Scheme and Perl are in this category. • A discussion of forms of identifiers: see the textbook, Section 5. 2. 5

Names and Identifiers 2 • Keywords or restricted words (e. g. if, var, type) provide the syntactic glue in a program. Consider: if (C) S 1 else S 2 • This is essentially a triple: C - S 1 - S 2. It might be equally well written as if C then S 1 else S 2 end if; or (C -> S 1 ; S 2) so long as the compiler and the programmer both understand the essential nature of this statement. • Similarly, while (C) S or while C do S is really a pair: C - S. 6

Names and Identifiers 3 • On the other hand, keywords and special words determine the style, the flavour of the language. • They help make programs readable (or not. . . ). • An aside remark: it is very easy to change keywords if it is a one-to-one change. We could, for example, trivially modify the lexical analyzer to get a compiler to recognize this: si C alors S 1 sinon S 2 fin si; instead of this: if C then S 1 else S 2 end if; 7

Variables • A variable in an imperative language, or an objectoriented language with an imperative basis, is a six-tuple: ‹name, address, value, type, lifetime, scope› • For example, if we write int x; we decide what will be the name and type of x. • The place of this declaration in the program decides where and how long x is available (scope, lifetime). Its address is determined when its program unit is executing, Finally, using x in statements decides what is its current value. 8

Variables • A problem in most programming languages: the same name can be reused in different contexts and denote different entities. For example: void a() { int b; /*. . . */ } float b() { char a; /*. . . */ } • Different types, addresses and values may be associated with such occurrences of a name. Each occurrence has a different lifetime (when and for how long is an entity created? ), and a different scope (where can the name be used? ). 9

Values of variables • The concept of value is more general: l-value (l for left) and r-value (r for right) are the source and target of an assignment. An example: x = y; • l-value is the address of x, r-value is the value of y. This becomes complicated for array elements that denote addresses which must be evaluated: T[i*2+1] = y; • This address depends on the current value of i. 10

Binding Points • An informal presentation • Binding times • Bindings of a variable • More on lifetime 11

Binding is not formally defined. Intuitively, we associate (that is, bind) an attribute with an entity. Examples of attributes are name, type, value. Binding occurs at various times in the life of a program. Three periods are usually considered: compile time (actually, translation time, because binding happens both in compilers and interpreters); load time (preparing object code for execution, pulling in the necessary code for built-in operations or operations defined in libraries of modules); run time (between starting the program's execution and its termination). 12

Static and dynamic binding We also distinguish two kinds of binding, depending on its duration: static binding is permanent during the life of the program; dynamic binding is in force during some part of the program's life. 13

Assigning properties to variables • variable name • compile time • described in declarations • variable address • load time or run time (e. g. C), • run time (e. g. Smalltalk) • this is usually done implicitly • variable type • compile time (e. g. Java), • run time (e. g. Scheme) • described in declarations, if bound at compile time 14

Assigning properties. . . (2) • variable value • run time, • load time (initialization) • specified in statements, mainly assignment • variable lifetime • compile time • described in declarations • variable scope • compile time • expressed by placement of declarations 15

Lifetime • Allocation of memory for an entity happens at load time or at run time. Two classes of variables are distinguished: • Static variables • Dynamic variables 16

Static variables • Memory for static variables is allocated once, before the program begins execution. • Fortran IV was an important language with such an allocation policy for all variables. This was inflexible, but it also was conceptually simple and inexpensive. Recursion was not possible. 17

Dynamic variables • Allocation is done after the program has started. • Two possibilities of dynamic allocation. • Explicit allocation and deallocation, that is, the programmer must do it. This is what we do when we use pointers; for example, in Pascal we allocate using new(p), deallocate using dispose(p). In C we use malloc(). • Implicit allocation (when a block is entered) and deallocation (when a block is exited). 18

Scope Points • Blocks and block structure • Anonymous blocks • Nesting diagrams • Call graphs • Static and dynamic scoping 19

Blocks and block structure • We group declarations and statements in order to • keep steps of a non-elementary activity together (e. g. , all steps of a sorting algorithm), • ensure proper interpretation of names. • Names are bound to various elements of a program. We refer to names in statements. • The scope of a name N means all places in the program where N denotes the same object. 20

Blocks and block structure (2) • Blocks can be nested. Names introduced in a block are called local bindings. A name referred to, but not declared, in a block must have been declared in a surrounding block. • Nesting of blocks is possible many languages, including C and its descendants. Nesting of procedures is allowed in Algol 60 and its descendants, but not in C. • A program, procedure or function consists of a heading and a named block. This is opposed to anonymous blocks, available in Algol 60, Algol 68, PL/I, Ada and C – but not in Pascal. 21

Anonymous blocks • An anonymous block is like a procedure defined (without a name) and immediately called once. • Such blocks are useful when a computation is required once and auxiliary variables are only needed in this computation. We do not want to declare them outside the block. (A procedure might as well be used to achieve the same effect. ) 22

Nesting diagrams Imagine C with nested functions. . . main P(){ int X; void A(){ char Y; void B(){ float Z; SB } SA } void C(){ int Z; SC } SP } P X A C A Y B B SB SA C SP Z Z SC 23

Call graphs P A C B Show which program units can call other units A loop "to itself" indicates possible recursion 24

Visibility, or the referencing environment P A SA C SP SC B SB P. X, P. A. Y, P. A. Y P. X, P. X P. C. Z P. A. B. Z Read U. V as "variable V declared in unit U". 25

Hole-in-scope (hiding): the same name is used in an enclosing block and a nested block. P SP X Y A B A SA X Z B Visibility. . . (2) Y Z SB The visible variables in SP : P. X, P. Y The visible variables in SA : P. A. X, P. Y, P. A. Z The visible variables in SB : P. X, P. B. Y, P. B. Z That is, P. X is visible everywhere in P except in SA, where P. A. X hides it and itself becomes visible. P. Y is visible everywhere in P except in SB, where P. B. Y hides it and itself becomes visible. There is no hole in the scope of P. A. Z or P. B. Z because they have disjoint areas of visibility. 26

Types of scoping • So far we saw static scoping (lexical scoping). It allows us to determine the use of every variable in a program statically, without executing it. • Dynamic scoping is an alternative with unclear advantages. The idea is to search for a name in a chain of called procedures, starting from the main program. This chain is built according to the visibility rules, but regardless of nesting. 27

Scoping example main P(){ int X; void A(){ X = X + 1; print(X); } void B(){ int X; X = 17; A(); } X = 23; B(); } The main program calls B, then B calls A. When A prints X, which X is it? With static scoping, it is X in P, the enclosing block of procedure A. The number printed will be 24. With dynamic scoping (search is done up the chain of calls), it is X in B, the most recently entered block with a declaration of X. The number printed will be 18. 28

Dynamic scoping • The dynamic scoping rule has been used in APL, SNOBOL-4 and in classic Lisp. It is not used much today. Even Common Lisp, a recent standard, has adopted static scoping. • Dynamic scoping is easier to implement than static scoping, but: • it is not possible to understand at compile time which variables refer to which objects, that is, type checking is impossible at compile time; • internal variables are not protected in the way we have come to expect: in our example X is local in B but A can nevertheless access it. 29

Constants, variable initialization Read Sections 5. 11 -5. 12 of the textbook. 30

Type checking Points • Operations and operands • Strong typing • Conversion and coercion 31

Types of operations Operations in programming languages normally require operands (arguments) of well-defined, specific types. Examples of operations • Boolean (comparison, conjunction, disjunction etc. ), • arithmetic (+, -, *, /, sin, tan, exp etc. ), • string (concatenation, substring etc. ), • assignment (this is an operation with two arguments), • passing a parameter to a procedure. 32

Type checking • Type checking ensures that an operator — when it is applied — gets arguments that it can handle. Type checking is done at compile time or at run time. • A type error occurs when an argument of an unexpected type is given to an operation. This too can be signalled at compile time or at run time. 33

Strong typing • A programming language has strong typing when all type errors can be discovered at compile time. This is not easy to achieve even in a strict language such as Pascal or Java. • Problem from Pascal: subtypes of enumerated types. type day = (mo, tu, we, th, fr, sa, su); workday = (mo, tu, we, th, fr); var x : day; y : workday; {. . . } y : = x; {? ? ? } 34

Strong typing (2) • A relaxed definition of strong typing: all errors can be discovered, preferably at compile time. • It is interesting that no popular programming language features perfect strong typing. There always are minor (or major) exceptions. Java is quite good on this count. 35

But… • Strict typing of operands is elegant and desirable, but it may be impractical. What happens if two operands make some sense together? • Example: float x; int n = 2; x = n * 3. 14; • This multiplication may be rejected by a strict compiler. We normally use an explicit conversion (also known as a type cast, as in Java): x = float(n) * 3. 14; 36

Type coercion • We can also rely on an implicit conversion called coercion: automatically change the form of an operand with a slightly inappropriate type. • For example, represent the value of n as a float and use this new (but equal) value in the multiplication. • C and Perl are the examples of languages with exaggerated use of coercion. 37

That's it This is all we can say in this course about names, binding, scope and type checking. There is a great deal more, and you can find those additional facts in the textbook, if only you are interested J. 38